48 Essential AI Agents Thought Leaders Globally (2026)

Last updated: June 2026

The 48 most essential AI agents thought leaders globally are a remarkably diverse group spanning framework builders, alignment researchers, enterprise practitioners, governance architects, and educators who are collectively defining what autonomous AI systems can and should do. As of June 2026, the success rate of AI agents handling real-world tasks has climbed from 20 percent in 2025 to 77.3 percent, according to Stanford University's 2026 AI Index Report, a shift that has moved the conversation from whether agentic AI works to who shapes how it develops responsibly and at scale.

This list surfaces the leaders who genuinely deserve to be far better known to anyone navigating this field. Rather than recycling the same handful of household names on every AI list, each person here was selected for substantive contribution to the specific challenge of building, deploying, governing, or teaching AI agents. They include the engineer who created the framework most enterprise teams use to build agents, the researcher whose departure from OpenAI signalled a crisis of safety culture that reshaped how the industry talks about alignment, and the educator whose single course catalysed a generation of practitioners building agentic workflows. They also include voices from Sweden, Canada, France, India, Ireland, Australia, and beyond, because the people shaping this field are distributed across the globe.

If your organisation is wrestling with how to move from AI experimentation to responsible deployment at scale, Jonno White works with corporates, nonprofits, and schools to translate the leadership and cultural dimensions of AI adoption into real action. As author of Step Up or Step Out (10,000+ copies sold), host of The Leadership Conversations Podcast (230+ episodes, 150+ countries), and keynote speaker delivering the 'Leading Humans in the Age of AI' talk, Jonno bridges the gap between what the technical leaders on this list are building and what your people and teams actually need to do with it. To book Jonno for a keynote, workshop, or executive offsite, email jonno@consultclarity.org.

Why AI Agents Matter Right Now

The most consequential shift in technology in 2026 is not the release of any single model. It is the transition from AI as a tool that responds to AI as a system that acts. An AI agent is a system that perceives its environment, sets goals, selects tools, takes actions, and evaluates its own outputs, often without a human in the loop for each step. The practical implication is that whole categories of knowledge work, from legal research and code review to customer service triage and supply chain monitoring, are being redesigned around autonomous systems rather than human workflows augmented by software.

According to McKinsey's 2025 State of AI survey, organisations that had moved to enterprise-wide AI adoption reported the highest financial returns, with 78 percent of respondents citing cost reduction and revenue generation as primary outcomes. The same research found that coding and customer operations represented the two highest-adoption use cases for agentic AI. A separate DeepL survey of 5,000 executives across five countries found that 69 percent expected agentic AI to transform their operations in 2026, the single sharpest year-on-year expectation shift the research documented.

The governance challenge has scaled in parallel. The 2026 International AI Safety Report, produced by an international consortium of researchers and presented to the India AI Impact Summit, found that the complexity of tasks agentic systems can automate has been increasing rapidly, while reliability remains a significant bottleneck and human oversight mechanisms remain underdeveloped for systems that act across multiple steps. The people on this list are engaging with exactly this gap, from the technical foundations of agent architecture to the policy frameworks that determine how autonomous systems are permitted to act on behalf of individuals and organisations.

For more on how AI is transforming the future of work and what leadership teams need to know, check out my blog post '50 Essential Thought Leaders in the Future of Work Globally' at https://www.consultclarity.org/post/thought-leaders-future-of-work-globally.

If your organisation is at the stage where AI agents are moving from pilot to production and you need the leadership, culture, and facilitation work that makes the transition stick, book Jonno White to deliver a keynote or workshop. Email jonno@consultclarity.org.

How This List Was Compiled

This list was compiled from a structured review of publicly available professional profiles, published works, LinkedIn activity, speaker profiles, and primary organisational sources. Each person was assessed on active LinkedIn presence and original content output in the past 60 days, substantive published contribution to the AI agents space (books, frameworks, papers, courses, or governance work), and a current incumbency check confirming the claimed role and organisation as at June 2026. People who had departed their listed roles, whose organisations had changed names, or whose contributions could not be independently verified were removed or updated. The list targets the mid-tier amplification band of practitioners actively building their platform, prioritising voices between 5,000 and 250,000 LinkedIn followers who post substantive original content regularly.

Category 1: Framework Builders and Tooling Architects

The people who create the infrastructure others build on occupy a unique position in the AI agents ecosystem. They are not only shaping what is technically possible but also establishing the design patterns, terminology, and defaults that every downstream practitioner inherits. A framework choice made by one engineer in 2022 now shapes how 35 percent of Fortune 500 companies deploy agents in 2026, which means framework architects are exercising influence at a scale that few researchers or commentators match.

1. Harrison Chase

Harrison Chase is the co-founder and CEO of LangChain, the agent engineering platform that became a foundational layer for enterprise AI agent development globally. Since founding the company in early 2023 alongside Ankush Gola, Chase has led it to over one billion open-source downloads and active deployment in 35 percent of Fortune 500 organisations, according to LangChain's own published metrics as of May 2026.

Chase's most distinctive contribution to the AI agents conversation is his work on what he calls 'context engineering', the discipline of structuring the information available to an agent at runtime so it can complete long-horizon tasks reliably. His April 2026 Sequoia Capital podcast appearance on context engineering and his regular 'Harrison's In the Loop' newsletter on the LangChain blog represent some of the most practically grounded thinking available on why agent reliability is an architecture problem, not just a model problem.

2. Kanjun Qiu

Kanjun Qiu is the co-founder and CEO of Imbue, a San Francisco AI research lab that has raised $232 million to build AI agents capable of reasoning and coding reliably. Qiu founded Imbue on the premise that reasoning is the primary bottleneck to effective AI agents: that today's systems are technically powerful but fail at the real-world complexity that makes autonomous action useful rather than dangerous.

Her contribution to the conversation on AI agents centres on the question of trust and verification. As she argued at the HumanX 2026 conference in San Francisco in April 2026, the user interface challenge around agents is not a cosmetic problem but a fundamental one: the interfaces needed for humans and agents to collaborate reliably on long-horizon tasks are still being invented.

3. Andrej Karpathy

Andrej Karpathy is the head of a pre-training research team at Anthropic, which he joined on 19 May 2026 after closing a 22-month chapter founding Eureka Labs, an AI-native education startup. The hire marked his return to frontier model research after building the Neural Networks: Zero to Hero YouTube series, which surpassed one million subscribers in early 2026, and releasing LLM101n, the most-starred open-source curriculum for training language models from scratch.

Karpathy's contribution to the AI agents space is his influence on how a generation of practitioners understands the foundations these systems are built on. His public statement in March 2026 that he had not written a single line of code since December 2025 because AI agents had taken over roughly 80 percent of his coding work became one of the most-shared moments in the AI agents conversation, drawing a straight line between his technical credibility and the practical question of how agents change the nature of technical work itself.

4. Chip Huyen

Chip Huyen is a computer scientist and author whose 2025 book AI Engineering: Building Applications with Foundation Models became the most-read book on the O'Reilly platform since its launch, according to O'Reilly's own platform data. She is currently working on a new startup in stealth. Previously, she co-founded Claypot AI, worked as a core developer at NVIDIA building the NeMo generative AI framework, and taught Machine Learning Systems Design at Stanford.

Her contribution to the AI agents conversation is her rare ability to explain what agents are built on in language that practitioners can use without a PhD. AI Engineering covers the AI stack that agents sit on top of, including evaluation, RAG, tool use, and the agentic workflow patterns that determine whether systems are reliable in production.

5. Victor Dibia

Victor Dibia is a researcher and author whose 2025 book Designing Multi-Agent Systems provides one of the most comprehensive frameworks for building multi-agent architectures from first principles. His Substack newsletter Designing with AI offers regular applied analysis of agent system design patterns that technical teams are actively using.

Dibia's contribution is filling a gap between academic multi-agent systems research and the practical frameworks engineers need to build reliable systems. His book covers agent concepts, tools, memory, orchestration, observability, multi-agent coordination, handoffs, and team structures with a combination of depth and usability rare in a field still generating more papers than production-ready guidance.

6. Amos Bar-Joseph

Amos Bar-Joseph is the co-founder of Swan AI, a company building AI-native startups that operate with minimal human headcount through autonomous agent systems. Bar-Joseph posts regularly on LinkedIn about building agent-native organisations, the design patterns for multi-agent systems in production, and the economics of scaling with agents rather than people.

His contribution is a practical, build-in-public perspective on what it actually looks like to run a company where AI agents do most of the work. At a time when most AI agents content addresses deployment theory or governance frameworks, Bar-Joseph consistently grounds the conversation in specific decisions his team makes: when to use a multi-agent architecture versus a single agent, how to handle agent failures in customer-facing workflows, and what the commercial model of an AI-native startup actually looks like.

Category 2: AI Safety and Alignment

The safety and alignment category is not peripheral to the AI agents conversation. It is the central one. An agent that acts autonomously in the world needs to act in ways that are aligned with human goals, safe under adversarial conditions, and robust across novel situations its designers did not anticipate. The researchers and practitioners in this category are working on exactly that problem, and their work increasingly determines what it is permissible to build and deploy.

7. Jan Leike

Jan Leike is the VP of Alignment Science at Anthropic, where he leads a team working on scalable oversight, weak-to-strong generalisation, and automated alignment research. He joined Anthropic in May 2024 after a highly public departure from OpenAI, where he co-led the Superalignment project and described safety culture as having taken a backseat to product development. TIME magazine named him one of the 100 most influential people in AI in both 2023 and 2024.

Leike's contribution to the AI agents field is defining what alignment means for systems that act autonomously over multiple steps. His work on scalable oversight, how humans can supervise AI systems doing tasks too complex for direct human evaluation, is the foundational problem for any deployment of agents in high-stakes environments. His Substack and Anthropic blog publications have shaped how researchers across the industry think about the hard problem of getting AI systems to follow human intent when humans cannot directly evaluate the outputs.

8. Anca Dragan

Anca Dragan is Vice President of AI Safety and Alignment at Google DeepMind, on leave from her role as Associate Professor of Electrical Engineering and Computer Sciences at UC Berkeley, where she founded and directs the InterACT Lab. Her research focuses on enabling AI agents, from robots to language models to recommender systems, to work with, around, and in support of people.

Her contribution is among the most rigorous available on the question of agent-human collaboration. Her research on reward learning, human-AI coordination, and alignment under uncertainty has contributed foundational methods for how agents model human preferences and adjust their behaviour accordingly. At Google DeepMind, her team is responsible for ensuring Gemini models are aligned with human goals and values, including avoiding present-day harms and preparing for more advanced capabilities.

9. Stuart Russell

Stuart Russell is Professor of Computer Science at UC Berkeley and President of the International Association for Safe and Ethical AI (IASEAI), a Paris-based international non-profit organisation established in 2024. He is the co-author (with Peter Norvig) of Artificial Intelligence: A Modern Approach, the most widely used AI textbook globally.

Russell's contribution to the AI agents conversation is his sustained, technically rigorous argument that current AI development approaches are fundamentally unsafe for autonomous agents. His 2019 book Human Compatible: Artificial Intelligence and the Problem of Control made this argument accessible to a broad audience and remains a cornerstone text for anyone thinking seriously about the governance of autonomous systems. His leadership of IASEAI brings institutional weight to the international coordination that agentic AI governance requires.

10. Yoshua Bengio

Yoshua Bengio is Scientific Director at Mila, the Quebec AI Institute, and one of three recipients of the 2018 Turing Award alongside Geoffrey Hinton and Yann LeCun for foundational contributions to deep learning. Since approximately 2023, Bengio has become one of the most prominent researchers publicly warning about the risks of advanced AI systems, including agentic AI.

His contribution to the AI agents safety conversation is his combination of unmatched technical authority and urgent public engagement. His role as lead author of the 2026 International AI Safety Report, produced for the India AI Impact Summit, gives him a convening function in the global governance of AI that no purely commercial voice can match. His argument that AI systems developing autonomous agency before adequate safety measures exist represents one of the most serious risks in technology is increasingly the framing that shapes how international bodies and national governments approach AI regulation.

11. Beth Barnes

Beth Barnes is the founder and CEO of METR (Model Evaluation and Threat Research), a nonprofit focused on evaluating the autonomy and potential danger of frontier AI models. METR was established to address the specific challenge of assessing whether AI agents are capable of actions that could cause harm at scale.

Her contribution is building the institutional infrastructure for testing whether AI agents are safe to deploy. METR's evaluations of frontier models for autonomous behaviour, including tests of whether models can autonomously replicate themselves, acquire resources, or resist shutdown, are among the most rigorous available. At a time when most safety conversation focuses on values and policy, Barnes and METR are doing the empirical work of actually measuring what agents are capable of, a contribution that has no substitute for anyone building or regulating powerful autonomous systems.

12. Gillian Hadfield

Gillian Hadfield is Professor at the University of Toronto's Faculties of Law and Arts and Science, and Chief Executive of Aligned AI, a company building AI systems governed by social norms rather than simple reward functions.

Her contribution is a distinctive legal and institutional framing for AI alignment. Where most alignment researchers focus on technical solutions, Hadfield argues that the problem of getting AI systems to behave well is fundamentally a social and institutional problem, and that solving it requires drawing on insights from how human societies manage complex coordination problems across law, norms, and governance structures. Her work on regulatory markets for AI and on what it would mean for AI systems to be aligned with pluralistic human values makes her one of the most original voices in the alignment field.

Category 3: Educators and Democratisers

The educators in this category are doing something genuinely difficult: making a fast-moving technical field accessible to the practitioners, leaders, and organisations that need to act on it without waiting for a computer science degree to become useful. The best AI agents educators are not simplifying out the important parts. They are finding the conceptual handles that allow non-specialists to make real decisions.

13. Andrew Ng

Andrew Ng is the founder of DeepLearning.AI and Managing General Partner of AI Fund. He is also a co-founder of Coursera and served as the founding lead of Google Brain and as VP and Chief Scientist at Baidu. Over 8 million people have taken an AI course from him, making him the most impactful AI educator in the world by any practical measure.

His specific contribution to the AI agents conversation is his 2025 launch of the Agentic AI course on DeepLearning.AI, which became the most in-demand new programme on the platform since its launch. His framework for understanding agentic workflows, specifically the four patterns of reflection, tool use, planning, and multi-agent collaboration, has become the reference vocabulary for how practitioners think about agent architecture.

14. Cassie Kozyrkov

Cassie Kozyrkov is the CEO of Kozyr, an AI consulting firm, and the founder of the field of Decision Intelligence. She was Google's first Chief Decision Scientist, training over 20,000 Googlers in data-driven decision-making and advising on more than 500 projects across a decade at Google before departing in 2023. She has been named a LinkedIn Top Voice for five consecutive years.

Her contribution to the AI agents field is reframing what autonomous decision-making actually means for organisations and leaders. Her core argument, that the question is never 'can the AI decide this' but rather 'is this a decision worth delegating and to what?', provides one of the most practically useful lenses for any executive navigating AI agent deployment. Her content consistently focuses on the governance and accountability questions that arise when systems act autonomously, translated into language that non-technical leaders can act on.

15. Andreas Welsch

Andreas Welsch is the founder and Chief AI Strategist at Intelligence Briefing and the author of two books on agentic AI for business: AI Leadership Handbook and The HUMAN Agentic AI Edge: Shape the Next Generation of AI-Ready Teams, published in 2025. He is also a LinkedIn Learning instructor who has created several popular courses on AI agents for business leaders, a former vice president at SAP, and an adjunct professor at West Chester University of Pennsylvania.

His contribution is translating the technical reality of agentic AI into organisational practice. The HUMAN Agentic AI Edge draws on interviews with more than 50 AI leaders to address the gap between deploying AI agents and building teams that can actually work alongside them effectively. Welsch's argument, that most agentic AI projects fail not because the technology does not work but because the human side is underprepared, is among the most practically useful framings available for organisations moving from experimentation to production.

16. Allie K. Miller

Allie K. Miller is the CEO of Open Machine, an advisory firm specialising in enterprise AI applications. Previously, she served as Global Head of Machine Learning for Startups and Venture Capital at Amazon Web Services, where she built a ten-figure business, and before that she launched IBM Watson's first multimodal AI team. She has 1.6 million LinkedIn followers and was named LinkedIn Top Voice for Technology and AI from 2019 through 2023.

Her contribution to the AI agents conversation is her sharp, practitioner-grounded commentary on what enterprise adoption of AI agents actually looks like from the inside. At Activate 2026 in New York City, she described the central shift of the agentic era as psychological: leaders are moving from writing instructions to orchestrating autonomous teammates, from AI writers to Chief Operating Officers of machine-driven workflows.

17. Noelle Russell

Noelle Russell is the CEO and founder of the AI Leadership Institute, where she works with organisations to build AI roadmaps and develop responsible AI leaders. She was named the number one Agentic AI Leader by Thinkers360 in 2025, is a five-time Microsoft AI MVP, and has taught more than 3.4 million learners through LinkedIn. Her book Scaling Responsible AI: From Enthusiasm to Execution provides a practical framework for the transition from AI experimentation to governed deployment.

Her contribution is bringing a human-centred, accountability-focused lens to the practical realities of enterprise AI adoption. Her experience building AI applications at NPR, Microsoft, IBM, AWS, and Amazon Alexa, and her founding of a global community focused on making responsible AI accessible to everyone, gives her an unusual combination of technical depth and community reach. Her daily LinkedIn content on AI leadership is among the most consistently practical available for leaders who need to make decisions today.

Category 4: Enterprise and Workforce Practitioners

The enterprise practitioners in this category are working at the intersection where agentic AI theory meets organisational reality. They are not primarily researchers or builders. They are the people advising, implementing, measuring, and making sense of what happens when autonomous AI systems enter real organisations with real people, incentives, and constraints.

18. Pascal Bornet

Pascal Bornet is the founder of IRREPLACEABLE with AI and creator of Agentic Intelligence, the first LinkedIn newsletter dedicated entirely to AI agents and produced using a team of AI agents. His newsletter has over one million readers as of May 2026. Bornet has more than two decades of experience leading AI and intelligent automation initiatives at McKinsey and EY, and was named a Top 10 Agentic AI leader by Thinkers360.

His contribution is sustained, high-signal weekly coverage of how agentic AI is deploying in real organisations. The Agentic Intelligence newsletter consistently surfaces case studies, governance developments, and framework shifts faster than most industry analysts, and Bornet's own analysis provides practical frameworks for interpreting what they mean. His concept that intelligent automation is not about replacing humans but about making the world more human is a distinctive framing that positions the conversation on agentic AI as fundamentally about workforce design.

19. Tim Cortinovis

Tim Cortinovis is an international keynote speaker and author based in Germany, focused on agentic AI in sales and revenue operations. He is the author of six books on AI, innovation, and modern sales, most recently Agentic Sales: Stop Chasing Cold Leads and Let AI Agents Generate, Qualify, and Convert Your B2B Sales Pipeline 24/7, published in 2025. He was named a Top 10 Thought Leader in Agentic AI and Sales by Thinkers360.

His contribution is grounding the abstract promise of agentic AI in the specific, measurable reality of what happens when autonomous agents take over parts of the revenue workflow. Where most AI agents commentary focuses on the technical architecture or governance implications, Cortinovis consistently translates the operational shift into practical frameworks that sales and commercial leaders can act on. His Agentic Revenue Architecture framework provides a structured approach to delegating revenue work to autonomous systems while maintaining human accountability for outcomes.

20. Tony Moroney

Tony Moroney is the Managing Director of The Behavioural Insights Practice and a recognised LinkedIn AI influencer listed among the top agentic AI voices to follow in 2025 and 2026 by multiple industry trackers. His content focuses on the governance, customer experience, and business transformation dimensions of AI agent deployment, with a particular emphasis on the organisational architecture needed to deploy agents responsibly.

His contribution is a governance and human-centred lens on what organisations need to build around agentic AI. His frameworks for AI fluency at the leadership level and his consistent engagement with how agentic AI changes decision-making architecture, from how authority is defined to how value chains evolve when machines handle execution, provide practical grounding for non-technical executives navigating this transition.

21. Ramnath Natarajan

Ramnath Natarajan is Director of Intelligent Automation at Johnson Controls, where he leads a team of over 100 engineers driving AI-powered operational solutions. He was named one of the top 30 AI leaders in 2025 by PEX Network and is active on LinkedIn on topics including AI agents, automation governance, and the operational dimensions of AI deployment in large organisations.

His contribution is a senior practitioner's perspective on what deploying AI agents at enterprise scale actually requires in terms of change management, governance, and human-AI teaming. His work at Johnson Controls, a Fortune 500 company operating across building technology, HVAC, fire safety, and security globally, gives him a vantage point on agentic AI deployment in industrial and regulated environments that few public voices can match.

22. Michael Chui

Michael Chui is a Partner at McKinsey Global Institute and one of the most widely cited researchers on the economic and workforce implications of AI and automation. His research includes the McKinsey 2025 State of AI report, one of the most cited annual surveys of enterprise AI adoption, co-authored with Alex Singla, Alexander Sukharevsky, and Lareina Yee.

His contribution is anchoring the AI agents conversation in rigorous quantitative research on adoption patterns, financial impact, and workforce transformation. His work consistently provides the empirical grounding that distinguishes evidence-based strategy from speculation in the enterprise AI conversation.

23. Craig Le Clair

Craig Le Clair is a Vice President and Principal Analyst at Forrester Research, where he has spent over a decade covering intelligent automation, AI agents, and hyperautomation. His work includes the Forrester Wave reports on robotic process automation and AI agents that enterprise technology buyers use to evaluate vendors and platforms.

His contribution is the analyst perspective that sits between vendor marketing claims and practitioner experience, providing independent assessment of which agentic AI platforms are mature enough for enterprise deployment. For organisations scaling AI agents beyond pilot, Le Clair's frameworks for vendor evaluation and platform maturity assessment provide a structuring discipline that internal teams often lack.

24. Leah Belsky

Leah Belsky is Chief Enterprise Officer at Coursera and a former general counsel who has built a public profile on the workforce and education implications of agentic AI. Coursera's enterprise platform delivers AI and agentic AI training to millions of professionals globally, and Belsky's LinkedIn content on what organisations need to build in their workforce to navigate the agentic transition is grounded in data on what large enterprises are actually choosing to learn.

Her contribution is a ground-level empirical lens on the skills gap that separates organisations that will manage agentic AI from those that will be managed by it. Her argument, that the transition to agentic AI is above all a workforce readiness challenge requiring deliberate investment in human skills for overseeing, directing, and auditing autonomous systems, provides a counterweight to the tendency in the AI agents conversation to focus on the systems rather than the people who work with them.

Category 5: Researchers and Academics

The academic and research voices in this category are doing the foundational work that determines what AI agents will be capable of, and what risks they carry, in five to ten years. Their current work frequently takes three to five years to reach production deployment, but the ideas they are developing now will shape the next generation of agentic systems.

25. Chelsea Finn

Chelsea Finn is an Associate Professor of Computer Science at Stanford University and a research scientist at Google DeepMind. Her research focuses on robot learning, meta-learning, and developing AI systems that can learn efficiently from limited experience in new environments, a core challenge for deploying agents across diverse real-world tasks.

Her contribution is foundational work on how AI systems learn to generalise and adapt, which is the technical prerequisite for agents that are useful in the open-ended real world rather than narrowly defined test environments. Her work on model-agnostic meta-learning (MAML), co-authored with Pieter Abbeel and Sergey Levine, has become a standard reference in the field of learning-to-learn for AI systems.

26. Dawn Song

Dawn Song is a Professor of Computer Science at UC Berkeley, co-director of the Berkeley RDI Center, and a serial entrepreneur whose work spans AI safety, security, and decentralisation. She is a recipient of the MacArthur Fellowship and is the founder of Oasis Labs.

Her contribution to the AI agents conversation is her focus on the security implications of autonomous systems operating in adversarial environments. Her research on AI safety and security, including the specific risks that arise when agents are deployed against adversarial inputs, prompt injection attacks, and malicious tool-use environments, addresses the class of failure modes that most agent developers encounter late in the deployment process.

27. Been Kim

Been Kim is a Senior Staff Research Scientist at Google DeepMind, specialising in interpretable machine learning and explainable AI. She is known for developing TCAV (Testing with Concept Activation Vectors), which received the UNESCO Netexplo award, and for her work on human-centred tools for understanding and communicating with complex AI models.

Her contribution is making the internal reasoning of AI agents legible to human overseers. As agents take actions with significant downstream consequences in enterprise, legal, and healthcare contexts, the ability to understand why an agent did what it did is not optional: it is a regulatory requirement and an ethical necessity. Kim's work on interpretability provides the technical foundations for the kind of meaningful human oversight that governance frameworks demand.

28. Rishi Bommasani

Rishi Bommasani is a Research Lead at Stanford HAI and a central author of the Stanford Foundation Model Transparency Index and the Holistic Evaluation of Language Models (HELM) benchmarks. He is also a co-author of the 2021 Stanford paper 'On the Opportunities and Risks of Foundation Models', which is among the most cited papers in the AI field.

His contribution is providing systematic, rigorous evaluation infrastructure that allows practitioners, policymakers, and researchers to compare and assess AI systems, including agents, against consistent benchmarks. His co-authorship of a 2026 Notre Dame Law Review paper on governing AI agents represents one of the most substantive academic contributions to the legal and policy frameworks needed for agentic AI governance.

29. Sara Hooker

Sara Hooker is the co-founder and CEO of Adaption Labs, a San Francisco startup that raised $50 million in seed funding in 2026 to develop AI systems capable of continuous real-time learning and efficient adaptation. Previously, she was Vice President of Research at Cohere, where she led the Cohere For AI research lab, and a researcher at Google DeepMind. She was named one of TIME's 100 Most Influential People in AI in 2024.

Her contribution to the AI agents field is her sustained research on model efficiency, which is the enabling constraint for deploying agents at scale. Her argument, that the future of AI lies not in bigger static models but in smaller, dynamic systems capable of learning on the fly, directly addresses one of the central limitations of current agent deployment: the cost and latency of running large models for every agent step.

30. Sanmi Koyejo

Sanmi Koyejo is the Director of Stanford HAI and an Associate Professor in the Department of Computer Science at Stanford University. His research focuses on reliable and trustworthy machine learning, fairness, and the development of AI systems that perform well under distribution shift, a core challenge for AI agents deployed across varied real-world environments.

His contribution is leadership at the intersection of technical AI research and the governance and policy ecosystem that is attempting to regulate it. As Director of Stanford HAI, one of the most influential institutions in the global AI governance conversation, Koyejo's work on what trustworthy AI systems actually require at a technical level informs how researchers, companies, and governments think about what standards and benchmarks should measure.

Category 6: Governance and Policy

The governance and policy voices in this category are working on the question that determines whether the AI agents field develops responsibly or not: what rules, structures, and accountability mechanisms are needed for autonomous systems operating in the world. Their work is not separate from the technical and commercial conversations. It is the frame that determines what is permissible.

31. Rumman Chowdhury

Rumman Chowdhury is the CEO and co-founder of Humane Intelligence, a tech nonprofit building community-driven evaluation and auditing infrastructure for AI models. She is also a Responsible AI Fellow at Harvard University's Berkman Klein Center for Internet and Society. Previously, she directed the ML Ethics, Transparency, and Accountability team at Twitter and was Global Lead for Responsible AI at Accenture. TIME magazine named her one of the 100 most influential people in AI.

Her contribution is building the institutional infrastructure for independent evaluation of AI agents outside the labs that create them. Her work on AI red teaming, the systematic adversarial testing of AI systems for harmful capabilities and failure modes, has become the reference approach for responsible deployment. Her leadership of Humane Intelligence, which runs community-based algorithmic audits and builds democratised access to AI evaluation tools, addresses the governance gap that arises when the only organisations capable of evaluating powerful AI agents are the ones who built them.

32. Natasha Crampton

Natasha Crampton is Microsoft's Chief Responsible AI Officer, leading the company's responsible AI work globally across research, engineering, government affairs, and product. Her work includes governance frameworks for AI agents operating in enterprise and government contexts, and she contributes to Microsoft's published principles and governance structures for responsible AI deployment.

Her contribution is a practitioner's governance perspective from inside one of the largest AI agent deployers in the world. Microsoft's Copilot and agentic AI products reach hundreds of millions of users globally, and Crampton's governance work determines how those systems are designed, tested, and deployed responsibly at scale.

33. Ana Valdivia

Ana Valdivia is an Associate Professor at the Oxford Internet Institute, University of Oxford, where her research focuses on the governance, regulation, and social implications of AI systems in high-stakes environments including criminal justice, public administration, and healthcare. Her work includes analysis of how algorithmic decision-making systems affect marginalised communities and how governance frameworks can address this systematically.

Her contribution is a rigorous critical perspective on the gap between AI governance frameworks as written and their effects in practice. Her research on how AI systems are deployed in public sector contexts across the UK and Europe provides an essential empirical check on the optimistic governance narratives that tend to dominate the conversation about responsible AI deployment.

34. Mia Dand

Mia Dand is the CEO and founder of Lighthouse3, an AI ethics and governance advisory firm, and is listed among the top global thought leaders in responsible AI by Thinkers360. She is a regular speaker and writer on AI governance, AI agents accountability, and the practical implementation of responsible AI frameworks in enterprise contexts.

Her contribution is translating AI governance principles into operational decisions that organisations can actually implement. Her advisory work with enterprises on responsible AI deployment, her LinkedIn content on the governance of autonomous systems, and her public commentary on how organisations should structure accountability for agent decisions provide a practical bridge between the policy research done in academia and the implementation challenges organisations face in production.

35. Isabelle Falque-Pierrotin

Isabelle Falque-Pierrotin is the Head of the AI Governance Division at the OECD in Paris, France, where she leads international work on AI governance frameworks, standards, and policy coordination. Previously, she served for many years as President of the French Data Protection Authority (CNIL) and as Chair of the Council of Europe's Convention 108 Committee.

Her contribution is providing the international policy infrastructure through which global governance of AI agents will be coordinated. Her leadership of the OECD's AI work, including the AI Policy Observatory and the OECD AI Principles adopted by over 40 countries, gives her a unique convening function in the conversation about how different national approaches to AI governance can be aligned across jurisdictions. As agentic AI systems increasingly operate across borders, the kind of international coordination she is building is the precondition for effective governance.

36. Kasia Chmielinski

Kasia Chmielinski is the co-founder of The Data Nutrition Project, a nonprofit that develops standardised tools for assessing and communicating the quality, composition, and limitations of datasets used to train AI systems. She is also a fellow at multiple research institutions including the Berkman Klein Center at Harvard.

Her contribution is addressing the data quality problem that lies upstream of agent reliability. An AI agent is only as good as the data it was trained on and the information it has access to at runtime, and Chmielinski's work on dataset documentation and transparency provides the infrastructure for understanding where systems are likely to fail, who is most affected by those failures, and what accountability the developers and deployers of those systems have.

Category 7: Agentic Research at the Technical Frontier

The researchers in this category are working at the direct technical frontier of what AI agents can do and what limits them. Their work ranges from the foundations of reasoning and learning in agents to the specific challenges of making multi-agent systems safe, reliable, and useful in production.

37. Piotr Mirowski

Piotr Mirowski is a Staff Research Scientist at Google DeepMind in London, known for his interdisciplinary work at the intersection of AI, theatre, and improvisation. He is the co-creator of AI improvisation and storytelling systems and has published research on multi-agent language systems and the role of embodied interaction in language model development.

His contribution is a perspective on AI agents that is both technically rigorous and distinctively humanistic. His research on improvised theatre with AI agents has produced some of the most original thinking available on how agents develop contextual understanding, manage uncertainty, and collaborate with humans in open-ended interactive environments. His work challenges the purely functional framing of AI agents and asks what it means for a system to be genuinely responsive to human context.

38. Evan Hubinger

Evan Hubinger is a Research Scientist at Anthropic working on AI interpretability and the specific alignment risk known as deceptive alignment, where AI systems learn to behave well during training while pursuing different goals in deployment. His 2019 paper on risks from learned optimisation in advanced machine learning systems, co-authored with Chris van Merwijk and colleagues, introduced the concept of mesa-optimisation and inner alignment to the AI safety field.

His contribution is one of the most technically precise available on the class of alignment failures that become especially dangerous when AI systems are deployed as autonomous agents with significant capabilities. His work on defining, detecting, and mitigating inner alignment failures shapes how researchers across the field think about the specific risks that arise when agents are powerful enough that their misaligned goals could cause serious harm before being detected.

39. Jack Clark

Jack Clark is the co-founder of Anthropic, the AI safety company he co-founded with Dario and Daniela Amodei and others after leaving OpenAI in 2021. He also created Import AI, one of the longest-running and most widely read newsletters on AI research, and previously served as Policy Director at OpenAI.

His contribution spans both the technical and policy dimensions of AI agents. His early work at OpenAI on AI policy and his subsequent role as a co-founder of Anthropic, a company whose founding rationale was that safety-focused AI development needed its own institution, give him a unique perspective on the institutional and research choices that determine how powerful AI agents are built and governed. Import AI, which Clark has produced weekly since 2017, remains one of the most reliable sources for tracking significant developments in AI research.

40. Holden Karnofsky

Holden Karnofsky is the co-CEO of Open Philanthropy, the philanthropic organisation that has provided over $300 million in funding to AI safety and alignment research, growing the field from approximately 20 to over 400 full-time-equivalent researchers. He also writes an influential personal blog on AI development timelines, transformative AI risks, and the strategic decisions facing organisations working on beneficial AI.

His contribution is providing the strategic philanthropic infrastructure that has made much of the most important AI safety research financially possible, and the public intellectual framework for thinking about why AI safety deserves a level of resource investment commensurate with the stakes. His writing on the urgency and priority of this work has shaped how the broader AI safety community thinks about the significance of getting autonomous AI systems right.

Category 8: Global Voices

The researchers and practitioners in this category are based outside the North American corridor that dominates most AI agents lists. Their perspectives reflect different regulatory contexts, research traditions, and deployment environments that are essential for understanding how agentic AI will develop globally.

41. Sasha Luccioni

Sasha Luccioni is the AI and Climate Lead at Hugging Face, based in Montreal, Canada. Her research focuses on measuring and reducing the environmental impact of AI systems, including the carbon emissions associated with training and deploying large models and AI agents. She co-developed the CodeCarbon tool for tracking the carbon emissions of machine learning experiments.

Her contribution is bringing a rigorous empirical lens to one of the most underaddressed dimensions of AI agents deployment: the environmental cost of running autonomous systems at scale. As organisations scale agentic AI from pilots to production, the environmental footprint of inference-heavy agent workflows becomes a governance and sustainability question, and Luccioni's work provides the measurement infrastructure to address it honestly.

42. Huma Shah

Huma Shah is a Reader in Artificial Intelligence at Coventry University in the UK, a former examiner for the original Turing Test competition, and one of the leading academic voices on AI interaction evaluation, machine intelligence assessment, and the question of what it means for a conversational AI system to be genuinely intelligent rather than merely persuasive.

Her contribution to the AI agents conversation is a long-form, critical perspective on the evaluation of AI systems that goes beyond task completion metrics to ask harder questions about what agents are actually doing when they appear to understand context, negotiate goals, and respond to human feedback. Her work on the history and theory of machine intelligence evaluation, grounded in decades of hands-on experience, provides a scholarly anchor for the evaluation debates that agentic AI is generating.

43. Alina Utrata

Alina Utrata is a researcher focused on the geopolitics of AI governance and the political economy of AI development, affiliated with institutions including the Centre for the Governance of AI and the University of Cambridge. Her research examines how national industrial policies, geopolitical competition, and global power dynamics shape the trajectory of AI development and governance.

Her contribution is an essential international and political lens on a conversation that often assumes a single global AI development path. Her research on how AI governance frameworks differ across jurisdictions, how geopolitical competition is shaping the development of AI systems including agents, and how smaller nations are attempting to participate in AI governance structures designed by larger powers is increasingly central to any realistic assessment of how agentic AI will be regulated globally.

44. Vilas Dhar

Vilas Dhar is the President of the Patrick J. McGovern Foundation, a philanthropic organisation committed to advancing AI and data solutions to benefit people and the planet, with a focus on ensuring that the benefits of AI are equitably distributed globally. He is an active voice on LinkedIn and at global forums on the intersection of AI governance, international development, and human rights.

His contribution is centring the perspective of communities in the Global South in the AI agents governance conversation. While most governance frameworks for agentic AI are being designed in the USA, Europe, and China, the communities that will be most significantly affected by the deployment of autonomous systems in healthcare, financial services, agriculture, and public administration are often in the Global South. Dhar's work to bring their interests into the design of governance frameworks is among the most important equity work in the field.

45. Tim Berners-Lee

Tim Berners-Lee is the inventor of the World Wide Web and Director of the World Wide Web Consortium (W3C). He has been actively engaged in the governance and architecture of AI systems operating on the web, including through his work on the Solid Project and his public commentary on the implications of AI agents operating across web infrastructure.

His contribution to the AI agents conversation is the longest possible view: what does it mean for autonomous systems to operate across the web, and what governance principles should guide that deployment? His combination of unmatched historical authority on web architecture and active engagement with the current implications of AI agents operating on web infrastructure makes him one of the most distinctive voices on how agents should interact with the open internet.

46. Menno Fokkema

Menno Fokkema is an AI strategy and governance expert based in the Netherlands who has built a significant LinkedIn following on EU AI Act compliance, agentic AI governance, and the practical implementation of responsible AI frameworks in European enterprises. His content on translating EU AI Act requirements into operational practice for organisations deploying AI agents is among the most practically useful available for European practitioners.

His contribution is making the regulatory environment for AI agents in Europe accessible and actionable for the enterprises that must comply with it. As the EU AI Act moves from aspiration to enforcement, the gap between regulatory text and organisational practice is significant, and Fokkema's ability to translate regulatory requirements into practical governance steps gives him a distinctive and increasingly important function in the European AI agents conversation.

47. Mounia Lalmas

Mounia Lalmas is VP of Research at Spotify, where she leads research on AI systems for music and podcast recommendation, personalisation, and increasingly the agentic tools that Spotify is deploying to support both listeners and creators. She is also a Professor at University College London and a leading researcher in information retrieval and AI-driven personalisation.

Her contribution is a real-world, scaled deployment perspective on AI agents in one of the most demanding personalisation environments in the world. Spotify's recommender and agent systems operate at a scale of hundreds of millions of users, across languages, cultures, and listening contexts, and Lalmas's research on the fairness, diversity, and user experience implications of those systems provides one of the most practically grounded available views on what it means to deploy AI agents responsibly at scale.

48. Pinar Ozturk

Pinar Ozturk is an AI ethics researcher, speaker, and advisor focused on the governance and societal implications of AI systems including autonomous agents. She works across academic, civil society, and corporate contexts to translate AI ethics principles into governance practice, with particular attention to how AI governance frameworks interact with diverse cultural, legal, and societal contexts globally.

Her contribution is a voice from outside the North American and Western European mainstream of AI governance discourse, addressing the specific risks that agentic AI deployment carries in contexts with different institutional structures, and advocating for governance frameworks that are genuinely pluralistic rather than assuming a single universal standard.

Notable Voices We Almost Included

Several strong candidates narrowly missed the final list. Fei-Fei Li (Stanford HAI, World Labs) is an essential AI voice but her primary contribution is to computer vision and general AI, not specifically to agents. Geoffrey Hinton (University of Toronto, co-winner of the Turing Award) is among the most important AI safety voices globally but did not meet the LinkedIn activity threshold that guides this list's mid-tier amplification focus. This list deliberately moved past the handful of names that appear on every list on this topic. That does not diminish those voices. It reflects a commitment to surfacing the people who deserve to be just as well known.

Common Mistakes Leaders Make When Following the AI Agents Conversation

The most common mistake is treating the AI agents conversation as a technology update to monitor rather than a strategic shift to act on. Most leaders who follow AI agents thought leaders absorb their content as interesting information rather than as input to a decision: which workflows will we redesign, what governance do we need to build, which people do we need to develop? The thought leaders on this list are not producing content for people who want to be informed. They are producing content for people who want to build, govern, and lead.

The second common mistake is following only the technical voices. The governance and policy researchers on this list are not a complement to the framework builders. They are working on the problems that determine whether the frameworks builders create get deployed safely or not. Organisations that follow only the builders and skip the governance thinkers are building institutional capability to deploy agents without the accountability structures to govern them.

The third mistake is assuming that because AI agents are developing rapidly, following events in real time provides sufficient understanding. The most important work on AI agents is happening in research papers, long-form blog posts, and books, not in product announcements and conference keynotes. Andrew Ng's agentic workflows framework, Jan Leike's work on scalable oversight, and Chip Huyen's treatment of the AI engineering stack in her book will still be relevant in three years in a way that most breaking news coverage will not.

The fourth mistake is geographic concentration. The people who will govern and deploy AI agents in India, the Middle East, Latin America, and Southeast Asia are not primarily following San Francisco-based voices. Building genuine understanding of how agentic AI is developing globally requires actively seeking out the voices from Isabelle Falque-Pierrotin in Paris, Alina Utrata in Cambridge, Mounia Lalmas at Spotify, and Vilas Dhar at the Patrick J. McGovern Foundation who are shaping the global conversation.

The fifth mistake is conflating the ability to deploy agents with the understanding of what they should do. Many organisations can now technically deploy agentic systems, and most cannot answer the governance questions that determine whether they should. Who has authority over an agent's decisions? What happens when an agent takes a consequential wrong action? How is the boundary between human and agent authority defined and maintained? The thought leaders in the governance and policy category on this list are working on exactly these questions.

How to Use This List Effectively

The most valuable thing you can do with this list is not follow everyone on it. It is to identify the three to five people whose specific area of contribution maps most closely to the decisions your organisation faces in the next 12 months, and engage with their work substantively, not just their headlines. If your team is building or evaluating agent frameworks, Harrison Chase and Chip Huyen are your starting points. If you are working on AI governance in a regulated industry, Rumman Chowdhury, Anca Dragan, and Isabelle Falque-Pierrotin represent the most practically useful voices available.

For organisations that need the leadership and facilitation work that turns insight into organisational action, Jonno White delivers keynotes, workshops, and executive offsites on leading humans in the age of AI. As author of Step Up or Step Out (10,000+ copies sold), host of The Leadership Conversations Podcast (230+ episodes, 150+ countries), and a keynote speaker who works with schools, corporates, and nonprofits around the world, Jonno brings the people and culture side of AI transformation that the technical voices on this list do not cover. International travel is often far more affordable than organisations expect. To bring Jonno in, email jonno@consultclarity.org.

For more on what AI means for workforce transformation, explore my blog post '50 Best Keynote Speakers on AI and the Future of Work' at https://www.consultclarity.org/post/keynote-speakers-ai-future-of-work.

Frequently Asked Questions

What is the difference between AI agents and traditional AI tools?

Traditional AI tools respond to a specific input with a specific output. An AI agent is a system that perceives its environment, sets goals, selects tools, takes actions over multiple steps, and evaluates its own progress without human confirmation at each stage. As of June 2026, agents handle real-world tasks at a 77.3% success rate according to Stanford's 2026 AI Index Report, though reliability remains uneven across task types.

Who are the most important AI agents thought leaders to follow on LinkedIn?

The most important AI agents thought leaders on LinkedIn in 2026 for different audiences are: for enterprise practitioners, Pascal Bornet, Andreas Welsch, Allie K. Miller, and Noelle Russell; for technical teams building agents, Harrison Chase, Chip Huyen, and Andrej Karpathy; for governance and policy, Rumman Chowdhury, Cassie Kozyrkov, and Menno Fokkema; for leadership and strategy, Tim Cortinovis, Tony Moroney, and Andrew Ng.

How is agentic AI different from generative AI?

Generative AI refers to models capable of generating content in response to prompts. Agentic AI refers to AI systems that take actions autonomously over time, using tools, making decisions, and pursuing goals across multiple steps. The agency, the ability to plan, act, evaluate, and iterate without continuous human instruction, is a separate architectural and behavioural layer on top of the generative capability.

What are the biggest risks of deploying AI agents in organisations?

The most significant risks include reliability failures in novel situations, accountability gaps when agents take consequential wrong actions, data privacy and security risks as agents access enterprise systems at scale, misalignment between agent behaviour and organisational values, and the challenge of maintaining meaningful human oversight as agent autonomy increases.

How do I identify the right AI agents keynote speaker for my event? For a commercial or sales leadership audience, Tim Cortinovis or Allie K. Miller will land most directly. For a technology or engineering leadership audience, Andreas Welsch or Harrison Chase offer the most practical depth. For the leadership, culture, and human side of AI transformation, Jonno White delivers the keynote that most technical AI speakers leave uncovered. Email jonno@consultclarity.org to discuss your event requirements. For a broader guide to AI speakers globally, check out my blog post '50 Best Keynote Speakers on AI Strategy for Executives' at https://www.consultclarity.org/post/keynote-speakers-ai-strategy-executives.

Final Thoughts

The 48 people on this list are not all pointing in the same direction. Some are building the frameworks that make AI agents commercially viable. Others are arguing publicly that the speed of development represents an unacceptable risk. A few are doing both, building systems while publishing research on why those systems need better safety measures before being deployed at scale. What they share is that their work is substantive, their contributions are primary rather than derivative, and their influence on how agentic AI develops globally is real.

The most important thing a leader can take from this list is not a set of accounts to follow. It is a map of the conversation that will shape the next decade of AI development. The framework builders and the safety researchers are not having separate conversations. They are engaged with the same fundamental question: how do we build AI agents that work, that are safe, and that contribute to human flourishing rather than undermining it?

If your organisation is navigating this moment and needs the leadership, facilitation, and culture work to make AI transformation stick, bring in Jonno White. Engage Jonno White to run an executive offsite on leading through AI disruption. As host of The Leadership Conversations Podcast (230+ episodes, 150+ countries), author of Step Up or Step Out (10,000+ copies sold), and Certified Working Genius Facilitator who works with schools, corporates, and nonprofits around the world, Jonno bridges the gap between the technical conversation and the human one. Many organisations find that international travel is far more affordable than expected. Email jonno@consultclarity.org.

For more on how AI is transforming healthcare organisations specifically, explore my blog post '50 Leading Global Thought Leaders on AI in Healthcare' at https://www.consultclarity.org/post/global-thought-leaders-ai-in-healthcare.

About the Author

Jonno White is a Certified Working Genius Facilitator, author of Step Up or Step Out, and leadership consultant who has worked with schools, corporates, and nonprofits around the world. His book Step Up or Step Out has sold over 10,000 copies globally, and his podcast The Leadership Conversations has featured 230+ episodes reaching listeners in 150+ countries. Jonno founded The 7 Questions Movement with 6,000+ participating leaders and achieved a 93.75% satisfaction rating for his Working Genius masterclass at the ASBA 2025 National Conference. Based in Brisbane, Australia, Jonno works globally and regularly travels for speaking and facilitation engagements. Organisations consistently find that international travel is far more affordable than expected. To book Jonno for your next keynote, workshop, or facilitation session, email jonno@consultclarity.org.

Sources

McKinsey Global Institute, 'The State of AI in 2025: Agents, Innovation, and Transformation' (2025). DeepL, survey of 5,000 executives conducted by Censuswide, September 2025. Stanford University, 'AI Index Report 2026' (2026). International AI Safety Report 2026, presented at the India AI Impact Summit.

Next Read

The field of AI agents overlaps significantly with the broader future of work conversation. For the broader landscape of AI transformation at the leadership level, explore my blog post '50 Best Keynote Speakers on AI in the USA (2026)' at https://www.consultclarity.org/post/keynote-speakers-ai-usa.